An architecture for Intrusion Detection based on an extension of the Method of Remaining Elements

Abstract

This paper introduces an Anomaly-based Intrusion Detection architecture based on behavioral traffic profiles created by using our enhanced version of the Method of Remaining Elements (MRE). This enhanced version includes: a redefinition of the exposure threshold through the entropy and cardinality of residual sequences, a dual characterization for two types of traffic slots, the introduction of the Anomaly Level Exposure (ALE) that gives a better quantification of anomalies for a given traffic slot and r-feature, an alternative support that extends its detection capabilities, and a new procedure to obtain the exposure threshold through an analysis of outliers on the training dataset. Regarding the original MRE, we incorporate the refinements outlined resulting in a reliable method, which gives an improved sensitivity to the detection of a broader range of attacks. The experiments were conducted on the MIT-DARPA dataset and also on an academic LAN by implementing real attacks. The results show that the proposed architecture is effective in early detection of intrusions, as well as some kind of attacks designed to bypass detection measures.